2 ms·
I guess Paul Christiano is the right person, given his old idea of "500 AI alignment scientists for every theoretical physicist". Other candidates probably don'
by Vecr 3y ago
I guess Paul Christiano is the right person, given his old idea of "500 AI alignment scientists for every theoretical physicist". Other candidates probably don't have enough experience directing large teams.
- randcraw 3y agoI'm pleased that NIST chose someone technical to head the work who is also well informed of the possible severe demerits of the technology. Christiano sounds like an ideal choice, so long as he can navigate the poisonous politics of Washington and the billions available to corporate lobbiests dedicated to pushing aside all obstacles to employing deep AI anywhere that promises profit.
- remarkEon 3y ago>AI alignment scientists What does an AI alignment scientist actually do? Because everything I've read about it makes me think it's just "make the AI do what I think it should do". Obviously, there are problems with that.